Factor subset selection for predicting sustainability of a telecentre using case-based reasoning

Background: Telecentre implementation in developing countries such as Ghana, has been inundated with failure. Various methods have been proposed to mitigate these failures, considering the substantial amounts invested. Various methods have been proposed. Recent work has shown that case-based reasoning (CBR) can be used to predict sustainability of a telecentre. Unfortunately, the factors that play a major role in the collapse of telecentres are yet to be determined. Objective: To combine two approaches (ICT4D evaluation and Machine Learning) and to identify the most important factors in the sustainability of a telecentre. Through the use of a case study, this paper applies the well-established feature subset selection (FSS) methodology to identify these influential factors. Method: We apply CBR with FSS on real life dataset to predict the Design Reality Gap score (DRGS). We compare two machine learning FSS methods with Expert Based Selection (EBS) method against benchmark using ArchANGEL. We also use the union and intersection of the selected factors to predict DRGS. Findings: We demonstrate through our experiments based on real world data sets that the combination of Machine Learning and Information Systems can identify these important factors. This gives a refreshing indication suggesting that it is feasible to use CBR with FSS to identify important factors of ICT initiatives and to adequately predict outcome of an initiative. Implication: Using this approach it is possible for managers and owners of telecentres to focus on the most important factors. This affords managers/owners an opportunity to channel limited resources to the most important factors thereby saving an ailing centre. © 2017 IEEE.